Artificial intelligence Video LDM from NVIDIA allows you to generate videos from short and simple text descriptions. Currently, the creation of short videos (4.7 s) in the mode up to 2048×1280 with 24 FPS is available.
The mechanism of Latent Diffusion Models allows you to create video from text without using powerful hardware. It uses a Stable Diffusion engine for its operation. The principle of LDM is to create a video from individual images (and train AI on them): a video is actually a sequence of generated pictures.
NVIDIA has posted many examples of Video LDM work on its site, including longer experimental videos. The public version of the neural network is not yet available, the company does not disclose the release date.
NVIDIA introduced hardware accelerators for ChatGPT, AI video processing, generative and other models
Source: NVIDIA